• No results found

Enhancements and Challenges in IEEE 802.11ah - A Sub-Gigahertz Wi-Fi for IoT Applications


Academic year: 2021

Share "Enhancements and Challenges in IEEE 802.11ah - A Sub-Gigahertz Wi-Fi for IoT Applications"


Loading.... (view fulltext now)

Full text





Master of Science Thesis

Examiners: Prof. Mikko Valkama Dr. Ali Hazmi

Examiner and topic approved by the Faculty Council of the Faculty of Computing and Electrical Engineering on 4th November 2015




MUHAMMAD QUTAB-UD-DIN: Enhancements and Challenges in IEEE 802.11ah - A Sub-Gigahertz Wi-Fi for IoT Applications

Tampere University of Technology Master of Science Thesis, 62 pages November 2015

Master's Degree Programme in Electrical Engineering Major: Wireless Communications Circuits and Systems Examiners: Prof. Mikko Valkama

Dr. Ali Hazmi

Keywords: M2M, IoT, Wi-Fi, IEEE 802.11ah, Restricted Access Window, Sectorization, Duty Cycle, Energy Eciency, Long Range

Internet of Things is a concept which brings ubiquitous connectivity to objects that we interact with in the course of our daily activities. With the projected estimates of the number of wireless connected devices reaching massive numbers, it is expected to revolutionize our daily lives signicantly. This sort of augmented connectivity will enable new applications in a myriad of domains including smart cities, smart houses, healthcare monitoring, industrial automation and smart metering. These applica-tions entail ecient operation of wireless networks with a large number of energy constrained devices. However, the existing infrastructure for wireless connectivity is not designed to handle such volume of projected growth.

Addressing this requirement, the IEEE 802.11ah task group is working on a new amendment of the IEEE 802.11 standard, suitable for high density WLAN networks in the sub 1 GHz band. It is expected to be the prevalent standard in many Internet of Things (IoT) and Machine to Machine (M2M) applications where it will sup-port long-range and energy-ecient communication in dense network environments. Therefore, signicant changes in the legacy 802.11 standards have been proposed to improve the network performance in high contention scenarios. In this thesis we evaluate the performance of many of the new features that have been introduced in the new standard including the Restricted Access Window, Sectorization and Sub-channel Selective Transmission mechanisms by means of analytical and simulated models. We propose novel Medium Access Control (MAC) layer algorithms which are shown to have improved the throughput and energy eciency performance in IEEE 802.11ah networks. We consider practical deployment scenarios in our simula-tions and evaluate the eects of challenges such as dense networks, interference from neighboring cells and duty cycle limitations on the performance metrics. Overall,


ii we nd that the advanced new features make 802.11ah standard a true IoT-enabling technology towards seamless integration of massive amount of connected devices in the future. Our research eort supports the notion that IEEE 802.11ah will be a key technology for future IoT and M2M applications especially in long-range and energy ecient deployments.




This thesis is written in partial fulllment of the requirements for a Masters of Science degree in Electrical Engineering at Tampere University of Technology, Tam-pere, Finland. This work was done under the Internet of Things program of DIG-ILE (Finnish Strategic Centre for Science, Technology and Innovation in the eld of ICT) which is owned by 40 companies in the Finnish industry and academia. The project was funded by the Finnish Funding Agency for Technology and Innovation (TEKES).

I would like to show my most sincere gratitude to Dr. Ali Hazmi who has been working with me tirelessly on this research eort. This work is the result of his ever present guidance, mentoring, ideas and most of all his patience throughout the course of this research. I would also like to thank Prof. Mikko Valkama who gave me the opportunity to work in his excellent research group and provided me with all the necessary support. His valuable suggestions and deep insight were the catalyst for this work's success. I would also like to thank my fellow researchers Orod Raeesi, Nader Daneshfar and Adnan Kiyani who were always eager to help me whenever I needed it.

I would also like to thank people from Ericsson Research, Finland including Behnam Badihi, Phelipe del Carpio and Parth Amin as they provided valuable comments and suggestions in this work. My special thanks goes to Atif Sajjad for reviewing this work. I would like to extend my deepest gratitude for my wonderful friends including Adnan, Zohaib, Angiras and Usama for helping me out on various occasions with this thesis.

I would like to thank my parents, Ameen and Kausar and my siblings, Abdullah and Abdulrehman whose constant support and aection have always been a motivating and inspiring factor for me. Without a doubt, I owe my achievements to them. Last but not the least, I would like to thank my dear wife, Marriam Irfan, whose love, patience and unconditional support has been the cornerstone of my success. It would not have been possible without her.

Tampere, 29.7.2015




1. Introduction . . . 1

2. Overview of IEEE 802.11ah . . . 4

2.1 Use cases and Requirements . . . 4

2.2 Challenges and Motivations for IEEE 802.11ah . . . 5

2.2.1 Dense Network Operation . . . 6

2.2.2 OBSS Problem . . . 6

2.3 Architecture of IEEE 802.11ah . . . 7

2.4 Physical Layer . . . 9

2.5 Medium Access Control Layer . . . 11

2.5.1 Channel Access . . . 11

2.5.2 Frame Shortening . . . 12

2.5.3 Virtual Carrier Sense . . . 14

2.5.4 Hierarchical Addressing and Page Slicing . . . 15

2.5.5 Target Wake Time . . . 16

2.5.6 Restricted Access Window . . . 16

2.5.7 Holding Schemes for Non-Cross Slot Boundary RAW . . . 18

2.5.8 Subchannel Selective Transmission . . . 20

2.5.9 Sectorization . . . 21

3. Simulation Settings and Platform . . . 23

3.1 Architecture of Omnet++ Platform . . . 23

3.2 Module Hierarchy . . . 24

3.3 Inter-module Communication . . . 25

3.4 Network Topology and Conguration . . . 26

3.5 Protocol Stack Implementation . . . 29

4. Analytical Considerations . . . 31

4.1 Accurate Analytical Model for Saturation Throughput of RAW . . . . 31



5. Performance Evaluations . . . 37

5.1 RAW Performance with Cross-Slot Boundary Condition . . . 37

5.2 Multi-Access Point Scenario . . . 41

5.2.1 TXOP Based Sectorization . . . 41

5.2.2 Subchannel Selective Transmission . . . 42

5.2.3 RAW and Group Sectorization . . . 43

6. Duty Cycle Challenges . . . 47

6.1 Theoretical Upper Limit of IEEE 802.11ah Duty Cycle . . . 48

6.1.1 STA Duty Cycle . . . 49

6.1.2 AP Duty Cycle . . . 51

6.2 IEEE802.11ah performance under duty cycle limits . . . 52

6.2.1 Saturated Trac Analysis . . . 52

6.2.2 Unsaturated Trac Analysis . . . 53

7. Conclusions . . . 57




2.1 Network Architecture of IEEE 802.11 WLAN . . . 7

2.2 Relay Architecture in IEEE 802.11ah . . . 8

2.3 Channelization for IEEE 802.11ah in (a) Europe and (b) US . . . 9

2.4 The proportion of the useful throughput for dierent MSDU sizes when MCS 0 and MCS 8 are used. . . 13

2.5 Restricted Access Window Structure . . . 17

2.6 Holding period description for dierent holding schemes . . . 18

2.7 BIN scheme for deterministic grouping in RAW slots withN = 10, NRAW = 2 and Ng = 5. . . 19

2.8 (a) RAW with grouping based on the AIDs (b) Group sectorization based on Spatial Distribution of STAs. Each color corresponds to a dierent slot/sector. . . 21

3.1 High level architecture of Omnet++ simulation platform . . . 24

3.2 Module hierarchy and network elements in Omnet++ . . . 25

3.3 NED Topology for a simple wireless network model . . . 26

3.4 TKENV view for network simulation . . . 27

3.5 Internal stack for a Wireless AP and Wireless STA . . . 29

4.1 Analytical model vs simulated results . . . 32

4.2 Simulated vs Analytical Saturated Throughput of RAW with Holding Period. µ = 10% . . . 34

4.3 Simulated vs Analytical Saturated Throughput of RAW without Hold-ing Period. . . 35



5.1 Distribution of stations around the AP . . . 38

5.2 Throughput for dierent number of RAW groups (N=1000) . . . 39

5.3 Energy Consumption for dierent number of RAW groups (N=1000) 40 5.4 Fairness for dierent number of RAW groups (N=1000) . . . 41

5.5 OBSS problem settings. . . 42

5.6 TXOP Based Sectorization . . . 43

5.7 Subchannel Selective Transmission . . . 44

5.8 Aggregate network throughput for two APs case. . . 45

5.9 Energy eciency for two APs case. . . 46

6.1 Typical data transmission in a simple IEEE 802.11ah network using the basic scheme: we assume uplink trac where the STA continu-ously (full buer case) sends DATA packets to the access point. . . . 49

6.2 Typical transmission in IEEE 802.11ah network . . . 52

6.3 Network Throughput in Saturated Trac for Limited Duty Cycle . . 53

6.4 Network Throughput in Unsaturated Trac with (black circles) and without (squares) the 2.8% Duty Cycle Limit . . . 54

6.5 Mean times within an hour for all stations to exhaust the 2.8% Duty Cycle Limit . . . 55 6.6 AP Duty Cycle in Unsaturated Trac with 2.8% Duty Cycle Limit . 56




2.1 IEEE 802.11ah use cases and related parameters . . . 5 2.2 Supported MCS and related parameters by IEEE 802.11ah for single

spatial stream and normal guard interval . . . 10 2.3 IEEE 802.11ah Minimum Receiver Sensitivity for 2MHz PPDU . . . . 11 2.4 IEEE 802.11ah Timing and Simulation Parameters (2 MHz mode) . . 14 5.1 Energy consumption for dierent transceiver modes . . . 40 6.1 Upper limit of duty cycle for a STA with dierent payload sizes for 1

MHz and 2 MHz channel widths . . . 50 6.2 Upper limit of duty cycle for AP with dierent uplink payload sizes

for 1 MHz and 2 MHz channel widths. Beacon size = 100 Bytes . . . 51 6.3 Trac Parameters for IEEE 802.11ah Use cases used in the simulation 53




IoT Internet of Things

RAW Restricted Access Window BSS Basic Service Set

MAC Medium Access Control

M2M Machine to Machine

WLAN Wireless Local Area Network

S1G Sub-1 Gigahertz

PHY Physical Layer

TXOP Transmit Opportunity

ETSI European Telecommunications Standards Institute OFDM Orthogonal Frequency Division Multiplexing

STA Station

AP Access Point

DS Distribution System

ESS Extended Service Set

ACK Acknowledgment Frame

NDP Null Data Packet

UL Uplink

DL Downlink

SST Subchannel Selective Transmission ISM Industrial, Scientic and Medical EDCA Enhanced Distributed Channel Access QoS Quality of Service




Recent years have seen a tremendous growth in the volume of wireless connected devices. It is expected that by 2020 the number of such devices will reach above 40 billion [1]. This increase can be attributed to various trends converging towards increased use of electronic devices for sensing, monitoring and reporting of real-time data. Such applications include smart cities, smart metering, medical monitoring, agricultural monitoring and industrial automation, just to name a few [2, 3]. The term Internet of Things was coined to describe such networks in which devices are able to share and report data without human intervention [4]. In a broad sense it refers to the interconnection of heterogeneous smart embedded devices within the internet infrastructure.

The IoT is a rapidly evolving concept and it is expected to transform our daily lives noticeably in the near future. In principle, the IoT will oer advanced connectivity of devices, systems and services. Several implementations of the IoT ideas are already being deployed or investigated, especially in areas like healthcare, industrial automation and smart cities, leading smoothly to the next stage of the IoT evolution. The future expansion of IoT and M2M communication particularly for sensor net-working is predicted to be enormous [5, 6, 7]. It is therefore of paramount importance to develop an infrastructure that can eciently support such numbers. In general, the IoT and M2M industries have dierent use cases and requirements, however their common denominators include low cost of devices and infrastructure, reliabil-ity, securreliabil-ity, connectivity to the internet and long battery operating times of end terminals.

Currently, several radio technologies exist that may be used for the M2M and IoT applications. These include technologies such as ZigBee [8], 6LoWPAN [9] or Blue-tooth [10]. In some cases, Wi-Fi or cellular networks have also been used for M2M applications. However, no explicit variant of these technologies exist which is opti-mized for IoT and M2M use cases or for sensor network purposes.


1. Introduction 2 Wireless Local Area Network (WLAN) communications and are among the most widely deployed solutions for the enterprise architecture. Although the major focus of many 802.11-related standardization eorts has been on delivering higher user throughput to a WLAN environment, there are multiple M2M-based scenarios that can benet from low-power device connectivity and relatively wider communication ranges when compared to the existing technology. Therefore, in order to overcome the above challenges and to provide an eective solution for IoT and M2M appli-cations, the Task Group ah (TGah) of the IEEE 802.11 standardization committee is working on the development of a new sub-1 GHz (S1G) amendment [11, 12, 13], namely the IEEE 802.11ah.

Studies regarding the deployment of the IEEE 802.11ah technology in IoT and M2M applications have substantiated its adequacy for the targeted use cases and the ability to operate well at the unlicensed Sub-1 GHz bands [14]. Studies done in [5, 15, 16] also establish IEEE 802.11ah as an ecient radio technology for M2M and IoT applications.

Compared to other IEEE 802.11 technologies and proprietary solutions like Blue-tooth and ZigBee, the IEEE 802.11ah can achieve higher ranges owing to its OFDM based PHY operating in the sub-1 GHz bands and lower data rates. Additionally, with the help of the newly introduced power saving mechanisms, IEEE 802.11ah can also noticeably reduce the energy consumption when compared to other exist-ing technologies and can support an increased number of devices per Basic Service Set (BSS) [15]. As the enhanced MAC features of the new standard make it suitable for high density and energy ecient WLANs, it holds great potential to be a catalyst for further market growth in the IoT and M2M communication spheres, including smart-homes, building automation, healthcare and other such applications [2, 3]. In this thesis, we present a comprehensive overview of the IEEE 802.11ah technology by describing the motivations behind its development, its main requirements and its general MAC and PHY characteristics. A key feature proposed in the IEEE 802.11ah standard is its Restricted Access Window (RAW) mechanism which enables the ecient operation of a large number of devices in a network. RAW mechanism reduces the collisions by allowing only a subset of stations in the network to contend for channel access periodically in their allocated time slots. We present an analytical framework to evaluate the saturation throughput performance of RAW mechanism. In this thesis we also extend the work of [17] by doing performance evaluation and enhancement of the RAW mechanism using dierent holding schemes that a station may adapt to prevent its transmission from crossing the boundary of its allocated RAW slot. Novel holding schemes as well as a grouping scheme for RAW


1. Introduction 3 are proposed and formulated and their performance is also evaluated.

Our other contributions in this thesis include primarily the performance evaluation and enhancement study of the most important newly proposed mechanisms in the IEEE 802.11ah MAC specication for throughput enhancement and interference mitigation, namely frame shortening, TXOP-based and Group Sectorization and subchannel selective transmission mechanisms. We have evaluated these features under practical deployment scenarios considering also the eect of nearby interfering BSSs by means of extensive system level simulations.

Additionally we have also investigated the eects of the maximum duty cycle limita-tion imposed by ETSI in Europe to prevent excessive emissions in the sub-gigahertz ISM frequency bands. If any chip maker is to sell its equipment for IEEE 802.11ah operation in Europe then they must comply with these regulations. Taking this into account, our work provides relevant and useful insight into the practical deployment considerations of this new standard.




The major design goal for IEEE 802.11ah is to fulll the requirement of many IoT and M2M applications. It implies that the standard has to be designed so that it supports the operation of a huge number of devices in contention based media under strict energy-constrained conditions. The IEEE 802.11ah promises to solve these problems by introducing several features and mechanisms which make it suitable for high density extended range wireless networks for battery powered end terminals [11].

Before we go on to explain the details of IEEE 802.11ah it is important to focus on the problems that it tries to solve and its possible and intended use cases. In the following section we briey describe these problems and hence the motivation behind the development of this new standard.

2.1 Use cases and Requirements

M2M and IoT systems are expected to enable a wide range of important services and applications including smart metering, healthcare monitoring, eet management and tracking, remote sensing, industrial automation, agricultural monitoring and on-demand business-charging transactions among many others [14, 18]. A massive number of devices are likely to be connected to enable these services and applica-tions which is the primary challenge for the existing technologies. Additionally, in many scenarios of the IoT use cases, end terminals have to operate without battery replacement or recharge for up to many years. Energy eciency is thus becoming of paramount concern when designing an M2M network [18, 19]. Furthermore, it is also deemed essential for the network operators to be able to oer M2M services and devices at lower cost levels while serving relatively larger areas.

Due to the increasing interest in IoT, M2M and sensor applications, the WLAN industry has also taken important steps to address this business segment by intro-ducing the new IEEE 802.11ah amendment [11, 20] to the IEEE 802.11 baseline standard mainly considering the above mentioned applications.


2.2. Challenges and Motivations for IEEE 802.11ah 5

Table 2.1 IEEE 802.11ah use cases and related parameters

Use Case STAs Data Rate Trac Type

Meter to Pole 6000 100 kbps C/P/B

Environmental Monitoring 300 100 kbps P/EB Industrial Automation 500 1 Mbps P(0.1s-100s)/B

Healthcare Systems 50 100 kbps P/EB

Extended Range WiFi 50 10 kbps B/Pmt

Cellular Ooading 50 20 kbps B

C = Continuous, P = Periodic, B = Burst, EB = Event Based, Pmt = Permanent

Use cases for the new amendment are divided into three broad categories namely [12]:

• Sensors and meters

• Backhaul sensor and meter data • Extended range Wi-Fi

Each of these use cases have dierent requirements for capacity, data-rates and trac intensity. The standard however contains provisions that can be adapted to satisfy the requirements for any of these use cases. Table 2.1 shows some of the intended use cases and their associated requirements for capacity, data rates and trac type. The new amendment mandates a minimum data rate of 100 kbps with a coverage radius of 1 km [13]. It is also a requisite to support at least one mode of operation capable of achieving a maximum aggregate Multi-Station data rate of 20 Mbps. Another requirement for IEEE 802.11ah is that it should support up to six thousands stations for outdoor applications and provide enhanced power saving mechanisms to support battery-powered operation with long replacement cycles [12, 21].

2.2 Challenges and Motivations for IEEE 802.11ah

The major challenge for the IEEE 802.11ah standard is the strict capacity and energy eciency requirements of many of the use cases. In the following we discuss these challenges and the proposed features in the standard to counter them.


2.2. Challenges and Motivations for IEEE 802.11ah 6

2.2.1 Dense Network Operation

As described in section 2.1, IEEE 802.11ah may require very dense operation of the network. For instance up to 6000 stations can be associated to a single Access Point (AP) in one of the use cases. It is not possible to provide acceptable data-rates in such high contention scenarios using the legacy MAC techniques. IEEE 802.11ah, therefore, describes many mechanisms to cope with this problem. Restricted Access Window (RAW) is one of the mechanisms which enables the operation of a large number of devices in a single BSS without degrading the throughput performance to inadequate levels. It does so by limiting the channel access in the BSS to a subgroup of stations in its assigned interval of time. Another mitigation mechanism described in the standard is the sectorization mechanism which counters issues such as inter-ference and hidden node problem in high density networks. These mechanisms are discussed later in detail.

2.2.2 OBSS Problem

The OBSS problem refers to the case when two or more BSSs, that are unsynchro-nized and operating at the same channel, are close enough to hear each other. In such a scenario, transmissions by some STAs in one BSS will be heard by the STAs in the other one, and eventually degrade the overall performance. Such a study was done in [5] where the authors concluded that the OBSS problem degrades the performance substantially as the overlap between the two access points is increased. To get a handle on this, the IEEE 802.11ah specications provide the sectorization mechanism. The sectorization mechanism not only helps reduce the interference from adjacent BSSs but also gets rid of the hidden node problem which is also a profound source of impairment in wireless networks using basic access mechanism. There are two kinds of sectorization mechanisms described in the new standard which we will cover later in detail. Subchannel selective transmission is another tech-nique that can be used to reduce the impact of interference from neighboring BSS. In this thesis we thoroughly evaluate the performance of these novel mechanisms through extensive simulations which give a realistic estimate of the performance improvements achievable by using these features.

Now that we have described the motivations behind this new standard and the applicable usecases we can proceed with providing an overview of the standard itself.


2.3. Architecture of IEEE 802.11ah 7

Figure 2.1 Network Architecture of IEEE 802.11 WLAN

2.3 Architecture of IEEE 802.11ah

In an 802.11 network, a station (STA) is a single addressable unit which is the source or destination of a message. A STA can be xed, portable or mobile. The basic architectural unit of an IEEE 802.11 network is a Basic Service Set (BSS) which can be thought of as the coverage area in which the STAs remain connected to each other. The area covered by the BSS is termed as Basic Service Area (BSA). A STA outside the BSS can not directly communicate with the ones inside the BSS. An independent basic service set (IBSS) is formed when two or more STAs capable of communicating directly with each other operate in a BSS thus forming a so called ad-hoc network. In contrast to that, in an infrastructure BSS the STAs associate to a xed STA called the Access Point (AP) which broadcasts special management frames such as beacon frames to keep the network synchronized. Two or more BSS can be connected by means of a Distribution System (DS) to form an Extended Service Set (ESS). An AP acts as a gateway for connecting multiple BSS to form an arbitrarily large Wireless Local Area Network (WLAN). An ESS formed in this manner allows the associated STAs to communicate with each other via the APs and DS even if they don't belong to the same BSS. Fig. 2.1 shows the basic architecture of an IEEE 802.11 network. In addition to that the IEEE 802.11 networks also provide the possibility of forming mesh networks in a Mesh Basic Service Set (MBSS). In an MBSS each STA is connected to its neighbor STAs and multi-hop routing is used for delivery of packets while there is no central entity [22].

An IEEE 802.11ah network maintains the network architecture of the legacy IEEE 802.11 systems therefore the above mention network architectures are also incor-porated in the new standard. It supports xed, outdoor and point-to-multi-point applications while being compatible to the IEEE 802.11 management plane [13]. In


2.3. Architecture of IEEE 802.11ah 8

Figure 2.2 Relay Architecture in IEEE 802.11ah

addition to that it also supports a relay architecture in which the coverage area of an AP is increased by use of a two-hop relay mechanism. A high level abstraction of the relay operation is shown in Fig. 2.2. It can be seen in the gure that the relay consists of a relay-STA and a relay-AP. The relay-STA is connected to the Root AP and the relay-AP is connected to the end STAs. Frames can be transmitted between the Root AP and the end STAs using the Relays in both directions. Relays not only increase the coverage area of the BSS but can also decrease the transmission time (usually) and energy consumption for successful packet delivery.

In IEEE 802.11ah, dierent BSSs can be set up around the same or dierent carrier frequencies which are determined by the regulations in the region of action. Ad-ditionally these BSSs will operate on channel bandwidths that range from 1 MHz to 16 MHz depending on the channelization policy of the respective country. The channelization for IEEE 802.11ah in Europe and the US is shown in Fig. 2.3 [23]. Throughout this thesis we will consider only the infrastructure BSS in our analysis since it is the most commonly deployed network architecture for WLANs. In an IBSS the APs primarily broadcast management frames (e.g. beacon frames) which help the STAs to operate and remain synchronized within the BSS. APs are also in charge of setting up associations with the STAs entering the BSS. Additionally they serve data trac to STAs on the downlink and respond with ACKs for incoming uplink trac. Consequently, because of their key role in the BSS, the APs generally need to transmit more frequently than an individual STA.

In the following we describe the physical and medium access control layers of IEEE 802.11ah and some of the new features that the standard has introduced.


2.4. Physical Layer 9



Figure 2.3 Channelization for IEEE 802.11ah in (a) Europe and (b) US

2.4 Physical Layer

The draft standard IEEE 802.11ah proposes operation in the sub 1 GHz (S1G) license-exempt bands excluding the TV whitespace bands. The proposed bands for deployment in the current draft are 863-868 MHz in Europe, 916.5-927.5 MHz in Japan, 755-787 MHz in China, 917.5-923.5 MHz in South Korea, 866-869 MHz, 920-925 MHz in Singapore and 902-928 MHz in the U.S. The TV whitespace bands are instead targeted by the IEEE 802.11af standard. The standard includes several MAC enhancements and also provides mechanisms to allow coexistence with other systems operating in the same bands, namely IEEE 802.15.4 and IEEE P802.15.4g [24]. It species a communication range of up to 1 km and a minimum data rate of 100 kbps [11].

The IEEE 802.11ah PHY supports multiple space-time streams and multi-user trans-missions. It uses a downclocked version of the IEEE 802.11ac PHY and its primary design goals include multiple interoperable operation bandwidth modes and support for a wide range of data rates [19]. Consequently, 2MHz, 4MHz, 8MHz and 16 MHz channels are specied with an additional 1 MHz channel in order to achieve longer transmission ranges. The OFDM tone spacing across the dierent bandwidth modes is a constant 31.25 kHz.

It is mandatory for a station to support MCS 0 - MCS 2 for 1 MHz and 2 MHz channel bandwidths with a single spatial stream. For an access point (AP) it is mandatory to support MCS 0 to MCS 7 for all supported channel widths for a single spatial stream [11]. The MCS supported by the S1G PHY specied in IEEE 802.11ah for 1 MHz and 2 MHz channel widths are shown in Table 2.2. The corresponding coding rate (CR), data bits per OFDM symbol (DBPS) and data


2.4. Physical Layer 10

Table 2.2 Supported MCS and related parameters by IEEE 802.11ah for single spatial stream and normal guard interval

MCS Mod. CR 1 MHz 2 MHz DBPS DR (Mbps) DBPS DR (Mbps) MCS 0 BPSK 1/2 12 0.30 26 0.65 MCS 1 QPSK 1/2 24 0.60 52 1.30 MCS 2 QPSK 3/4 36 0.90 78 1.95 MCS 3 16-QAM 1/2 48 1.20 104 2.60 MCS 4 16-QAM 3/4 72 1.80 156 3.90 MCS 5 64-QAM 2/3 96 2.40 208 5.20 MCS 6 64-QAM 3/4 108 2.70 234 5.85 MCS 7 64-QAM 5/6 120 3.00 260 6.50 MCS 8 256-QAM 3/4 144 3.60 312 7.80 MCS 9 256-QAM 5/6 160 4.00 - -MCS 10 BPSK 1/4 6 0.15 -

-rate (DR) are also shown in the table.

In IEEE 802.11ah, transmissions are frame based with each frame consisting of multiple OFDM symbols. However during the design, the system was specically optimized in the lower bandwidth modes to support lower data rates and longer ranges that would be useful for power limited (e.g. battery operated sensors) devices operating in S1G bands.

The IEEE 802.11ah draft standard lists a few channel models including an outdoor and an indoor path loss model. In this thesis, we consider only the outdoor path loss model with macro deployment scenario where antenna height is assumed to be 15m above rooftop, as described as one essential 802.11ah deployment scenario in [25]. The path loss in dB is given by ( 2.1) where d is the distance in meters and carrier frequency is 900 MHz.

P L= 8 + 37.6×log10(d) (2.1)

The minimum receiver sensitivity for each MCS according to the standard is listed in Table 2.3.


2.5. Medium Access Control Layer 11

Table 2.3 IEEE 802.11ah Minimum Receiver Sensitivity for 2MHz PPDU

MCS Sensitivity (dBm) 0 -92 1 -89 2 -87 3 -84 4 -80 5 -76 6 -75 7 -74 8 -69 9 -67

2.5 Medium Access Control Layer

In order to address the IoT and M2M requirements the core changes and enhance-ments in IEEE 802.11ah are related to the MAC layer. The introduced MAC features aim mainly at enabling the IEEE 802.11ah technology to support the expected high number of stations with power limited characteristics while transmitting short pack-ets. In the following, we rst briey describe the essential MAC features and then focus on investigating the features addressing capacity requirements and interference mitigation.

2.5.1 Channel Access

The IEEE 802.11ah MAC provides Enhanced Distributed Channel Access services for an S1G STA using the services of the Distributed Coordination Function [11]. EDCA services mean that the STA in an IEEE 802.11ah network is a QoS station, supporting dierent priorities for dierent types of trac.

DCF is a type of Carrier Sense Multiple Access with Collision Avoidance (CSMA-CA) technique for channel access in a shared medium. In the DCF procedure, a station rst senses the channel to be idle for an interval equal to Distributed Interframe Space (DIFS). It then chooses a random number of backo slots from the range[0−CW]to wait before it can start its transmission. CW is the size of the contention window, which is initially CWmin. The receiving station rst checks if

the packet is destined for itself. If this is the case, then it sends an acknowledgment frame (ACK) back to the sending station after waiting for a period called Short Interframe Space (SIFS). SIFS is smaller than DIFS so ACKs have always higher priority than normal packets. A full transmission cycle therefore is the sum of


2.5. Medium Access Control Layer 12 DIFS, backo, duration of data frame, SIFS, ACK and the associated propagation delays. The sum of data frame duration, SIFS and ACK is also known as a transmit opportunity (TXOP).

If an ACK is not received by the sending station within the timeout interval, the packet is considered to be lost and is re-transmitted with the size of contention window doubled on each attempt until it reaches its maximum size CWmax. For

this reason, DCF is also termed as a binary exponential backo scheme. A packet is dropped after mlong number of unsuccessful transmission attempts. The contention

window is reset to CWmin if a packet is transmitted successfully which is indicated

by successful receipt of an ACK frame. A collision is said to have occurred when one or more stations transmit at the same time [22]. In legacy 802.11 networks, when a collision occurs, the station waits for Extended Interframe Space (EIFS) time before attempting another transmission, which is larger than DIFS to avoid possible collision again. However, in IEEE 802.11ah, EIFS is not used [26].

There are other coordination functions also available in IEEE 802.11ah such as Point Coordination Function (PCF) or Hybrid Coordination Function (HCF). We will, however, not discuss these as they are not mandatory to be supported and we only consider DCF in our work.

2.5.2 Frame Shortening

In the legacy IEEE 802.11 networks, MAC header overhead can even exceed 30% for a 100-byte payload, therefore, rendering the use of legacy frame formats impractical in sensor applications as it is typical to have frequent short payload transmissions and long transmission delays [12, 20]. Therefore, to reduce the control overhead, Task Group ah (TGah) has introduced short headers, short beacons and Null Data Packet (NDP) frames in the new standard.

The newly introduced short header excludes few uncritical elds like the Dura-tion/ID, the Quality of Service (QoS) and High Throughput (HT) elds. As a result, size of the MAC header can be reduced to as low as 14 Bytes in the new standard including the FCS eld. Additionally, to reduce channel occupancy time and subsequently power consumption, the IEEE 802.11ah introduces short beacon frames that are sent more frequently than normal ones and do not carry any redun-dant or uncritical information that can be alternatively obtained from the legacy beacons [26]. IEEE 802.11ah also proposes the use of Null Data Packets (NDP), a concept rst introduced in IEEE 802.11ac, in which packets have no payload from the MAC layer and contain all essential information in the PHY header. IEEE


2.5. Medium Access Control Layer 13

100 bytes0 256 bytes 512 bytes 0.2

0.4 0.6 0.8

MSDU size in bytes

P ro p o rt io n h MCS0

100 bytes0 256 bytes 512 bytes 0.05 0.1 0.15 0.2 0.25 0.3 0.35

MSDU size in bytes

P ro p o rt io n h MCS8 Legacy frame Short frame

Figure 2.4 The proportion of the useful throughput for dierent MSDU sizes when MCS 0 and MCS 8 are used.

802.11ah proposes the use of NDP frames for control frames such as ACK, CTS, BlockAck etc.

In order to assess the upper performance of the IEEE 802.11ah system, and to measure the throughput enhancement when short frames are considered, we assume that a single STA in an IEEE 802.11ah network successfully transmits a single data frame. The total transmission time can be divided in two parts, Ttr needed for the

MPDU transmission plus a constant overheadTOverhead [27] given in ( 2.2). Here we

have neglected the delay due to the propagation time.

T =TOverhead+Ttr (2.2)

where the constant overheadTOverhead is expressed as follows:


2.5. Medium Access Control Layer 14

Table 2.4 IEEE 802.11ah Timing and Simulation Parameters (2 MHz mode)

Parameter Description Value

Tsym OFDM Symbol Duration 40µs

PHY HDR Physical Layer Header Length 240µs

MAC HDR MAC Layer Header Length 14 B

ACK Acknowledgment 14 B

SlotTime The slot time 52µs

SIFS Short interframe space 160µs

DIFS DCF interframe space 264µs

CWmin Min. back-o window size in SlotTime 15

CWmax Max. back-o window size in SlotTime 1023

L Payload Size 256 B

mlong Maximum number of Tx attempts 4

δmax Propagation delay 6µs

Tg Guard Time (2×δmax) 12µs

TRAW Duration of RAW 1s

PT X Transmit Power 1mW

The frame transmission time Ttr varies according to the bit rate used by the STA.

If MCS 0 is used and if the frame size is of 256 bytes of payload (MPDU of total 290 bytes in normal frame case and of total 270 bytes in the shortened frame case), the proportion of the useful throughput η measured above the MAC layer in both cases [27] is given as follows:

η= Ttr


• Payload Size

Frame Size (2.4)

Given Table 2.4, proportion of the useful throughput is 0.35 and 0.55 in normal frame and short frame cases respectively. We can easily notice the considerable improvement of the achieved throughput when the shortened frames are considered. In Fig. 2.4 we show a comparison of the useful throughput for dierent MCSs and payloads sizes when legacy and short frames are considered. Here we also neglected the back-o time, as we assumed that only one STA is in the cell.

2.5.3 Virtual Carrier Sense

The legacy virtual carrier sense mechanism in 802.11 WLANs is called the Network Allocation Vector (NAV). It is used on top of physical carrier sensing to save power. Every station who is listening on the wireless medium can receive frames from other stations. When such a frame is detected which is not destined for the receiving


sta-2.5. Medium Access Control Layer 15 tion, that station reads the duration eld of the frame which indicates the amount of time this transmission will continue. It subsequently updates its Network Alloca-tion Vector (NAV) which is eectively a timer that indicates the time during which the network will be busy. The stations which set their NAV in this manner do not attempt to transmit until the NAV timer is zero which indicates that the channel is idle [28].

Since the duration eld is omitted when short MAC header is used in IEEE 802.11ah, NAV can not be used as the only virtual carrier sense mechanism. TGah has devel-oped a new carrier sense mechanism called response indication deferral (RID). By using RID it is possible to specify the type of response required by the transmitting station in the response indication eld of the PHY header. Based on this value the listening stations can defer channel access for an appropriate interval of time [26].

2.5.4 Hierarchical Addressing and Page Slicing

IEEE 802.11ah describes a hierarchical addressing scheme for the association iden-tiers (AID) of the stations in the infrastructure BSS. The hierarchy consists of 4 pages of 32 blocks each. Every block then consists of 8 sub-blocks, each having 8 stations. Therefore, a 13-bit AID is sucient to uniquely identify a station. In legacy 802.11 networks it was only possible to have a little over 2000 devices asso-ciated to an AP as the rest of the values in the 14-bit association identier (AID) were kept reserved. Also, the length of the Trac Indication Map (TIM) bitmap is increased from 2008 bits to 8192 bits [26] and hence up to 8192 stations can be addressed using this scheme within a single BSS.

This hierarchical addressing can facilitate the use of many other enhancement mech-anisms by allocating AIDs to stations with similar characteristics such that they fall into logical hierarchical groups. This type of addressing scheme may also help in reducing the size of the Trac Indication Map (TIM). Although the standard does not describe in detail how the TIM can be compressed, it does however provide a framework to do so using creative and ecient means and is an open research topic. An extension of this hierarchical structure is the page slicing mechanism. The page slice information element (PSIE) present in the DTIM beacon contains the information that whether there is buered data for at least one station in the page. The stations see that and calculate the approximate time at which they should wake up to receive the appropriate full beacon to retrieve buered data. This helps in conserving the energy for sensor stations.


2.5. Medium Access Control Layer 16

2.5.5 Target Wake Time

In an IEEE 802.11ah network a station can be either in Awake state where it is active or in the Doze state in which the station conserves battery by not listening to the channel transmissions. When a station goes to Doze state, it must notify the AP it is associated to, so that it can buer any frames for the dozing station. These buered frames can be sent to the dozing station by the AP upon receipt of a PS-Poll frame by the station. The Trac Indication Map (TIM) which is broadcasted as part of the beacon frame contains the information about presence of buered data for the stations. A station upon noticing that the AP has buered data for it, can send a PS-Poll frame and retrieve the data. For this purpose the dozing stations repeatedly need to wake up to listen to the beacon frames, at least in the legacy standard.

IEEE 802.11ah denes a new power saving mechanism called the Target Wake Time (TWT) specically for the stations which do not want to wake up to listen to each beacon frame. These stations can negotiate a wake up time with the AP at which they wake up for an interval called Service Period (SP) during which they can exchange frames. Stations can remain in doze state for extended periods of time using this mechanism thus increasing the battery life cycle of sensor nodes to months or even years.

2.5.6 Restricted Access Window

Restricted Access Window (RAW) is one of the most promising and well studied fea-ture of IEEE 802.11ah. A major portion of this thesis is also therefore dedicated to the performance analysis and enhancement of this feature. The importance of RAW mechanism is well established in [5] and [15] while [17] also describes the RAW oper-ation considering the cross slot boundary condition. RAW uses enhanced distributed channel access (EDCA) which uses DCF as the underlying access mechanism [11]. To support a large number of devices associated with a single AP, TGah has de-veloped a novel mechanism to reduce contention in the channel access. In this mechanism, during a particular time window called the restricted access window (RAW), a group of stations is allocated time slots during which they can contend for channel access. Stations are not allowed to contend for the channel outside their designated slots. To conserve power the remaining stations can sleep outside their RAW slots. Additionally, within a beacon interval there can be multiple RAWs as well as open access intervals during which any station is allowed to contend for the


2.5. Medium Access Control Layer 17

Figure 2.5 Restricted Access Window Structure

channel [11, 3]. Inside a RAW slot, stations in a group contend for the channel ac-cess using the DCF procedure, which is the same as used in legacy 802.11 networks and described in Section 2.5.1.

When a STA gets uplink data from the upper layers it can contend for the channel access at the beginning of its allocated RAW slot. It is however necessary that the STA is allowed channel access in that particular RAW slot. It stops attempting packet transmission as soon as the time assigned for that slot ends. It is also possible that a STA may not be allowed to use any of the RAW slots. In that case it may only attempt channel access in an open interval. Fig. 2.5 shows the structure of the time assignment in RAW during a beacon interval.

The grouping of stations may be done by the AP which assigns the group to a station by means of the RAW Parameter Set (RPS) element broadcasted in the beacon frame. Besides group assignment information, the RPS element contains essential control information about the RAW operation including duration of a slot and the number of RAW slots (NRAW). This information is used by the station to

calculate the total duration of the RAW (TRAW). A simple slot assignment criteria

is described in the standard in which stations are allocated RAW slots according to the mapping function dened as

i= (x+Nof f set) modNRAW (2.5)

Here,iis the index of the assigned RAW slot and ranges from 0 toNRAW−1andxis

the Association Identity (AID) of the station. Nof f set is a parameter that improves

fairness among the stations and is equal to the two least signicant bytes of the beacon FCS. If a total of N stations are present in the network then each RAW group has Ng =N/NRAW number of stations contending in a RAW slot.


param-2.5. Medium Access Control Layer 18

Figure 2.6 Holding period description for dierent holding schemes

eters during beacon interval so that either dierent STAs will use dierent RAWs or channel access parameters between RAWs are dierent. Additionally, the AP can congure periodic RAW so that channel access is possible for certain STA periodi-cally.

The RPS element also contains a eld called the cross slot boundary (CSB) condi-tion. If CSB condition is false i.e. the CSB eld is set to 0, then a station must not start a transmission if the remaining time in the current RAW slot, albeit non-zero, is not enough to complete the transmission before the end of the slot. Otherwise, if the CSB condition is true, it is allowed for a station to transmit even if the transmission will not be completed before the end of the current RAW slot [11]. In this thesis we consider only the non-cross slot boundary case and propose and analyze the performance of new schemes for backo counting within the holding period. We also propose a novel grouping scheme of stations for RAW based on the apriori information obtained from the backo states of the previous RAW. These schemes are discussed in Section 2.5.7.

2.5.7 Holding Schemes for Non-Cross Slot Boundary RAW

To prevent a transmission from crossing the RAW slot boundary, a station shall hold its transmission a certain amount of time before the end of the RAW slot, termed as the `holding period' [29]. No station is allowed to start a transmission within the holding period. Thus, it should at least be equal to the sum of DIFS and the time taken by one TXOP (Tdata+SIF S+Tack). It may also include a guard period to

2.5. Medium Access Control Layer 19

Figure 2.7 BIN scheme for deterministic grouping in RAW slots with

N = 10, NRAW = 2 and Ng = 5.

allow for propagation delays.

There can be several options for how the station counts its backo within the hold-ing period which we call here as a holdhold-ing scheme. The dierent holdhold-ing schemes analyzed in this thesis are described in this section. While `FIXED' and `DECRE-MENTING' schemes are mentioned in [17], it contains the analysis of only the `FIXED' holding scheme. In our work, we also evaluate the `DECREMENTING' holding scheme and propose and evaluate two new holding schemes which we term as `VARIABLE' and `HYBRID' schemes. We provide their detailed description in the following.

1. FIXED: In the xed holding period, the backo counter is frozen when the holding period starts. The station goes to idle and stores its backo state. The station restores its backo state in the following RAW.

2. DECREMENTING: In this scheme the station keeps decrementing its backo counter inside the holding period. If the counter reaches zero inside the holding period, it is renewed with the same contention window. Otherwise, the backo state is stored at the end of the slot boundary and restored in the next RAW. 3. VARIABLE: In this holding scheme, the station freezes its backo as soon as it generates the backo number and anticipates that the generated backo will cause the slot boundary to be crossed. This eectively decreases the useful slot duration but can be useful in some scenarios where using a larger backo or synonymously a larger holding period is benecial, as we will later see. 4. HYBRID: In this scheme the stations choose the holding scheme based on

their distance from the AP. The stations which are far from the access point choose the `VARIABLE' holding scheme and the ones which are near use the `FIXED' holding scheme.


2.5. Medium Access Control Layer 20 Fig. 2.6 shows the basic holding schemes described above. In addition to these hold-ing schemes we also propose and evaluate the performance of a new novel grouphold-ing scheme. After the end of the rst RAW within a beacon interval, all stations have their backo states stored. These states contain some information about the con-tention scenario present in the network and can thus be used to assign groups to stations in the next RAW so that channel utilization is increased. With the assump-tion that it is possible for the AP to know the backo states of all the staassump-tions, it can regroup them in each subsequent RAW in order to improve the network performance. Assuming that the AP has the backo states of all the nodes in the network, the new grouping scheme, which we term as the `BIN' scheme, works on the following algorithm. It rst sorts the list of backo counter numbers of all the stations in the network in ascending order, from which the index of thekth station in theith RAW group is found using mapping function in ( 2.6)

Ii,k =i+NRAW ×k (2.6)

where i ∈ [0−NRAW) and k ∈ [0−Ng). The procedure is illustrated in Fig. 2.7

forN = 10, NRAW = 2 andNg = 5. The procedure can be repeated after the end of

each RAW, whereas in the rst RAW the mapping function in ( 2.5) can be used. There can be a dierence of one between the number of stations in dierent RAW groups when group size Ng is not an integer.

The `BIN' scheme reduces the collisions by increasing the channel utilization as it places stations in a group with their backos in ascending order. It also reduces collisions by placing stations with equal backo counters in dierent groups and hence allows more packets to be transmitted in a RAW slot. In the following sections we evaluate the performance of the above mentioned schemes through analytical considerations as well as comprehensive network simulations.

2.5.8 Subchannel Selective Transmission

STAs in a BSS can benet from the fact that they may not require the whole available channel bandwidth for their transmissions. For instance, sensor stations may only support 1 MHz and 2 MHz channel bandwidths and can therefore select the best subchannel among the available channel bandwidth. IEEE 802.11ah denes the Subchannel Selective Transmission clause to enable this mechanism. An AP may indicate the support for SST by including an SST Operation Element in a management frame which contains the set of channels enabled for SST operation.


2.5. Medium Access Control Layer 21 1 6 15 9 11 5 3 2 4 8 7 12 10 16 14 13 1 6 15 9 11 5 3 2 4 8 7 12 10 16 14 13 (a) (b)

Figure 2.8 (a) RAW with grouping based on the AIDs (b) Group sectorization based on Spatial Distribution of STAs. Each color corresponds to a dierent slot/sector.

The AP may then include an SST Element in a beacon frame containing a subset of enabled SST channels on which SST operation is allowed in the following beacon interval.

The SST Element also contains a channel activity schedule containing information about expected DL and UL activity on the allowed subchannels and the duration for which the activity is allowed. In the downlink the AP may send sounding frames to the STAs on the allowed subchannels which in turn can be used by the STAs to estimate channel parameters in order to determine which subchannel is the best. Once a STA determines the best subchannel it can then use it in UL transmissions during the time indicated in the channel activity schedule. The AP may also use RAW for transmission of the sounding frames or it can send these frames periodically. The SST mechanism can help in mitigating selective fading and interference due to OBSS.

2.5.9 Sectorization

The new amendment proposes the use of Sectorization mechanism which addresses issues like channel contention, hidden node problem and interference from OBSS by partitioning the BSS into several spatial sectors. This partitioning is achieved by means of beamforming antennas to cover dierent sectors of the BSS. Depending on the considered deployment scenario (one AP or multi-APs), the IEEE 802.11ah spec-ications dene two types of sectorization mechanism namely Group sectorization


2.5. Medium Access Control Layer 22 and TXOP-based sectorization.

Group Sectorization

The group sectorization mechanism is developed with an intent to address the hidden node problem which is especially more pronounced in dense long-range networks. During the association procedure, the AP assigns each associated STA an ID of the sector which it belongs to. The AP divides the time into sector intervals and at the start of each sector interval it transmits a beamformed beacon to the sector which is allowed to access the channel during that interval. This alleviates the hidden node problem as all the stations belonging to the same sector can hear each others transmissions. The mechanism is very similar to RAW as illustrated in Fig. 2.8. In RAW mechanism, STAs in a given RAW slot are grouped based on their AIDs whereas in group sectorization the STAs in each others spatial vicinity are grouped into sectors.

TXOP-based Sectorization

TXOP based sectorization aims at minimizing the interference caused by OBSS while allowing the associated stations to transmit their data simultaneously. In TXOP-based sectorization, the AP starts a transmission with an omni beam which sets up the NAV of all listening stations. After the receipt of ACK from the destination node, the AP transmits the second frame such that part of it is omni and part is sectorized. If the OBSS STAs listen the omni beam but cannot listen the ACK nor the sectorized beam transmission, they can start their own transmission until the end of sectorized TXOP transmission. The standard also describes few other ways to implement the TXOP-based Sectorization. A comparative study of these schemes and their specic applications form an important research issue.

Having described the main features of the IEEE 802.11ah standard, in the follow-ing we rst describe our simulation platform and then develop an analytical model for saturation throughput performance of RAW. Since RAW and sectorization are fundamentally similar, the same model can be applied to calculate the saturation throughput of the sectorization mechanism. We then proceed with the detailed per-formance analysis of IEEE 802.11ah MAC features and our proposed enhancements under practical deployment scenarios by means of realistic system level simulations.




It is important to verify the result of network modeling by means of an analytical as well as simulated model. In this thesis we have used the Omnet++ tool for the network simulations for our analysis purposes. The accuracy of these simulations is then veried by means of two analytical models. In this section we will describe Omnet++ tool used for the system level simulations of our network model.

Omnet++ is an object-oriented discrete event network simulator based on the C++ language. It provides a rich set of tools and libraries to develop and simulate net-work components and protocols. It is an open source tool and therefore highly customizable according to specic requirements. It takes a modular approach on system design, allowing reusable, independent and highly ecient modules. The main components of the Omnet++ framework are described here.

3.1 Architecture of Omnet++ Platform

Fig. 3.1 presents the high level architecture and constituent blocks of the Omnet++ simulation platform. The arrows show the interaction between the high level mod-ules. SIM is the library linked to the simulations program and contains the classes which perform the tasks of a kernel for the simulations. It handles the I/O from the user, schedules the events and manages the execution of parallel simulations. It is also responsible for managing CPU resources.

There are two type of user interfaces available in Omnet++ namely CMDENV and TKENV. As the name suggests, CMDENV is a command line user interface which basically provides minimal information about the simulation events, the simulation progress and its conguration. This is a light user interface and does not require much computing resources. It is ideal to use when long and stable simulations are being run. The other user interface is TKENV which is a graphical user interface (GUI) and provides a rich set of tools to monitor, inspect and analyze the simulations in real time. It is a computing intensive interface and requires extra dependencies to be installed in order to run. However, this UI is well integrated to suit the needs of most Omnet++ users as it has advanced options for monitoring and debugging


3.2. Module Hierarchy 24

Figure 3.1 High level architecture of Omnet++ simulation platform

simulations. Its use is recommended when it is necessary to perform validation of a certain function in the simulations or when debugging is required in order to ensure correct behavior of the program. ENVIR is the library which contains the code common to both these interfaces.

In Object Oriented terminology, ENVIR is the interface while TKENV and CM-DENV are concrete implementations of this interface. The most important part of the ENVIR module is the `ev' facade object, which it presents to the Executing Model as well as the simulation kernel (SIM). The call to `ev' resolves at run-time to either CMDENV's or TKENV's implementation and it provides methods such as writing to the standard output/standard error and watching the value of a sim-ulation statistic in run-time. In addition to that ENVIR also contains the main() method which is the starting point of the simulation. This module is in fact in full control of the simulation and instructs the simulation kernel about which modules should be set up. It contains the main event loop as it invokes the simulation kernel for the required functionality. It also catches and handles any errors or exceptions that occur during the simulation.

Model Component Library (MCL) contains denitions for a wide range of frequently used modules in network simulations such as channels, networks, messages and com-pound modules etc. While MCL provides blueprints or classes for the functional blocks of simulation elements, the exact instances are held by the Executing Model. The information about these instances is also made available to the simulation kernel by the Executing Model.

3.2 Module Hierarchy

Omnet++ follows a simple module level hierarchy where `simple modules' make up the `compound modules'. In implementation terms, each simple module is a C++


3.3. Inter-module Communication 25

Figure 3.2 Module hierarchy and network elements in Omnet++

class which has a declaration and a denition, like in any other C++ implementa-tion. There is no limit on the levels of hierarchy or the number of simple modules that make up the compound modules. Modules interact with each other by passing `Messages'. Every module can have multiple `Gates' which act as the entry or exit point for receiving or sending messages to other modules. A gate is a directional entity and it can be used for incoming or outgoing messages or both. A simple dia-gram showing a network composed of simple and compound modules in Omnet++ and related entities is shown in Fig. 3.2.

This hierarchical approach to modules in Omnet++ allows the simulated networks to be modular and makes it trivial to implement any kind of protocol stack. The design goal of Omnet++'s simple modular hierarchy was to make the modules cohesive and decoupled. Cohesiveness implies that a module performs a well dened task by itself whereas decoupling means that it has no dependency on other modules. By following these best practices of object oriented design, Omnet++ makes it possible to simulate complex network scenarios while keeping the code base clean, reusable and easily maintainable with minimum eort. Usually the algorithms are implemented in simple modules while compound modules act as a container for a number of simple modules. Modules in Omnet++ can have several parameters which are used to congure it. The parameters can be specied in either the INI les or NED les which are described shortly after.

3.3 Inter-module Communication

Modules communicate with each other by means of exchanging `messages'. Usually these messages are sent through a module's `gates' which travel through a link or a predetermined path, but they can also be sent directly to another module. The


3.4. Network Topology and Conguration 26

Figure 3.3 NED Topology for a simple wireless network model

messages are dened in .msg les which follow a simple C-like syntax. Omnet++ comes with compiler support (MSGC) to generate C++ code from message denition les. This compiler generates appropriate getters and setters for the data structure elds used in messages which then can be used in module implementations.

The links or connections through which these messages travel can for instance rep-resent a channel with a path loss, delay, data rate and other related parameters. Otherwise, a link can simply be an interconnection of two modules within a single level of module hierarchy. The messages can be arbitrarily complex data struc-tures and may represent frames, packets, customers or jobs in a queue depending on the context. Omnet++ provides several built-in types for channels that can be customized by providing them appropriate parameters.

3.4 Network Topology and Conguration

In Omnet++, the structure of the network model is described in the NED (Network Description) language. The language as its name implies `describes' the composi-tion of the network as well as each individual component. The user can assemble simple and compound modules and connect their gates through links using the NED language. Basically it acts as a glue for dierent components in the network. The NED language supports object oriented features such as inheritance and interfaces which makes the components to be readily extensible and easily reusable.

To sum it up until this point the network topology is described using the NED language (.ned les), specic algorithms for active modules are written in C++


3.4. Network Topology and Conguration 27

Figure 3.4 TKENV view for network simulation

(.h/.cpp les) and message denitions are written in (.msg) les. Omnet++ requires every simple and compound module to have a NED declaration while its behavioral and algorithmic implementation is written in a C++ class registered with that NED declaration. Omnet++ provides us with a visual NED le editor for easy manipulation of network topologies. One such topology is shown in Fig. 3.3 which is the top level NED architecture of our basic simulation scenario. It contains an array of Access Points (APs) and Sensor Stations (STAs) which are congured so that stations are uniformly distributed around the APs in a radial region. This is shown in Fig. 3.4 which is the TKENV view when the simulation is run.

Such congurations can be provided by specifying the parameters of the modules either in NED les or in INI(initialization) les. INI les override the parameters already provided in the NED les and are usually used to provide common cong-urations for several modules. The source code of the simple NED topology shown above is given in Prog. 3.1 which has two parameters numHosts and numAPs. It also contains two compound modules named Wireless_STA and Wireless_AP which will be dened in their respective NED denitions. The parameters should be provided in the INI le in this example before executing the network model because NED les contains no default values for them.


3.4. Network Topology and Conguration 28

Program 3.1 NED Topology for a simple wireless network model

package wsn;

import inet . nodes . wireless . WirelessAP ;

import inet . linklayer . ieee80211 . Ieee80211NicSTA ;

network Sensor_Network { parameters: int numHosts ; int numAPs ; submodules:

sensor [ numHosts ]: Wireless_STA {

@display("r = , ,#707070 "); }

ap[ numAPs ]: Wireless_AP {

@display("p =61 ,124; r=,,# ff0000 "); }


A sample INI le for this network would look like Prog. 3.2 if we only consider the top level parameters. This would initialize the simulation with 10 wireless stations and 1 access point. The rst two lines here represent the name of the conguration and the name of the network.

Program 3.2 INI le for simple network model

[ General ]

network = Sensor_Network

**. numHosts = 10 **. numAPs = 1


3.5. Protocol Stack Implementation 29

Figure 3.5 Internal stack for a Wireless AP and Wireless STA

3.5 Protocol Stack Implementation

Until now we have described how basic networks are formed in Omnet++ using the NED language. However, the submodules that we have used in our previous example can be made arbitrarily complex to implement any kind of protocol stack. We have used the same blocks in this thesis to implement the physical and MAC layers of IEEE 802.11ah in Omnet++. On top of that there is a management module whose purpose is to manage the overall activity of a station including doze and awake times, beacon handling and packet handling from upper layers. Above the management layer, a trac generator module generates packets with specied size and inter arrival time (time between generation of consecutive packets). Moreover, every station has a mobility module which keeps track of the movement of the station and whether it can move or not. In this thesis we have only considered stationary nodes as mobility is not the prime focus of IEEE 802.11ah. There is also a battery module with every station which keeps track of the amount of energy consumed by it during the course of the simulation and other similar statistics. A diagram showing the internal structure of both a wireless station and wireless AP is shown in Fig. 3.5.


3.5. Protocol Stack Implementation 30 of the energy consumption statistics there. Therefore, no battery module is used with AP implementation. The AP receives the messages from all the stations in the network in uplink and after successfully decoding them, transmits an acknowledg-ment. It also needs to generate beacon frame periodically which are broadcasted to the whole network.




It is important to verify the results of the simulations by analytical means as they cannot be declared accurate with a good degree of condence without such ver-ication. In this thesis we have veried our work with two analytical models. More specically the analytical models presented in this chapter verify the satu-ration throughput of RAW and Group Sectorization mechanisms. The rst model is presented in [17] and gives an accurate model for saturation throughput of RAW mechanism. The second model present by us is a simple analytical model based on [30] which oers a good estimate of saturation throughput of RAW and veries our simulation results.

4.1 Accurate Analytical Model for Saturation Throughput of


The analytical model presented in [17] provides the saturation throughput for RAW under the basic `FIXED' holding scheme and ideal channel conditions. We use this model to prove that our simulations are accurate and henceforth analyze our newly proposed holding and grouping schemes described in section 2.5.7. The saturation throughput S is given as



×Etr×Ps, (4.1)

where L denotes the payload bits, NRAW is the number of RAW groups (slots),

TRAW is the duration of the RAW, Etr is the expected value of the number of

full transmission cycles and Ps is the probability of a station transmitting a packet

successfully. A full transmission cycle is the sum of DIFS, Backo and TXOP (DATA + SIFS + ACK).

In the foregoing analytical model, the probability of successful packet transmission is the same as mentioned by Bianchi in [30] and is given by ( 4.2). Here, Ng is the


4.1. Accurate Analytical Model for Saturation Throughput of RAW 32 2 4 8 16 32 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

Number of RAW groups

Throughput (Mbps)

32 STAs − ana 32 STAs − simu 64 STAs − ana 64 STAs − simu

Figure 4.1 Analytical model vs simulated results

RAW slot and τ is the probability that a station transmits in a randomly chosen slot time. The model assumes that the number of backo slots follows a geometric distribution between consecutive TXOPs in a RAW slot. Calculation of τ and Etr

along with rest of the analytical treatment can be found in [17].

Ps =


1−(1−τ)Ng . (4.2)

We use the Omnet++ tool for our simulations as described in Chapter 3. Fig. 4.1 presents a comparison of the aforementioned analytical model with our simulations using the `FIXED' holding scheme. The results are shown with 32 and 64 STAs associated to a single AP in the Infrastructure Basic Service Set (IBSS) with varying number of RAW groups. Saturated trac is considered which means that a station always has a packet to transmit. The eects of path loss and propagation delays are also ignored in this case and ideal conditions are assumed. Other system parameters are shown in Table 2.4. It can be seen that the analytical and simulated results match nearly perfectly with an error of about 3% or less.


4.2. Simple analytical model for Saturation Throughput of RAW 33

4.2 Simple analytical model for Saturation Throughput of


In this section we propose a simple analytical model for the RAW performance under saturated trac. Previously, the analysis work on IEEE 802.11 DCF perfor-mance either includes the Markov-chain-based analysis [30, 31, 32] or the mean value methodology [17, 33, 34, 35]. In both schemes it is assumed that the probability for transmitting a packet in an arbitrary slot is the same. We use the same assumption in our analysis.

To derive the analytical RAW performance, we consider a fully-connected IEEE 802.11ah network withN STAs accessing the wireless channel within a RAW period, i.e., there are no hidden terminals in the network. We also assume an ideal channel condition where no communication errors or capture eects occur. Furthermore, all transmitted packets are assumed to have the same length.

We start our analysis by noticing that within each RAW slot the stations are basi-cally contending for channel access using the standard DCF procedure. The main dierences are that the allocated time duration is smaller and the number of con-tending stations is fewer. Additionally in each RAW slot, there is a holding time period where the stations are not allowed to transmit due to the non-cross slot boundary feature imposed by the IEEE 802.11ah standard. In the following analy-sis we will rst assume that the whole raw slot is used, i.e. the holding time period is set to zero. Therefore the total throughput in RAW period is basically the aggregate throughput over allNRAW slots. In [17], a similar analytical study of the saturation

throughput in RAW has been considered where the mean value methodology has been used.

In our analysis we will extend the well referenced work by Bianchi [30], where he derived the saturation throughput performance of the IEEE 802.11 DCF using Markov-chain based analysis, to the RAW scenario. Let SDCF,N be the normalized

throughput as derived in [30] for DCF scheme withN STAs contending for channel access. We refer to this as the normalized Bianchi throughput and it is dened as the portion of the time utilized for successfully transmitting payload bits in a slot time. The expression for the normalized throughput is given as




, (4.3)


4.2. Simple analytical model for Saturation Throughput of RAW 34 0 20 40 60 80 100 120 140 160 180 200 220 240 260 280 300 320 340 360 380 400 Number of Stations Saturation Throughput (kbps) RAW, N RAW=2, ana

RAW, NRAW=2, sim RAW, NRAW=5, ana RAW, N

RAW=5, sim

RAW, NRAW=10, ana RAW, NRAW=10, sim

Figure 4.2 Simulated vs Analytical Saturated Throughput of RAW with Holding Period.

µ = 10%

respectively. E[P]is the average payload size andσis the duration of an empty time slot. Finally, Ts and Tc represent the time taken by a successful and failed (due to

collision) transmission respectively. More details on how to obtain each individual parameter can be found in [30].

If the total RAW duration isT seconds, then the Bianchi saturatedST


through-put in


Related documents

The recent proliferation of research on virtual fashion design and online retail presence compelled us to immerse our physical selves and investigate online shopping in Second

“ I had a lot, lot more to concentrate on… didn’t really need to know who I was because I just had so many things to do like my health and I didn’t really have time to think

Abstract: Integrated circuits have emplane the era of a system-on-chip (SoC), which advent integrating all peripheral electronic devices into a single chip. In the proposed

The results of amino acid sequence homology analysis showed that on the carA operon there homologous 218 amino acids, 7 amino acids are not homologous, and the

Qualcomm Snapdragon, Nvidia Tegra are among the most widely used mobile processors.. 3.1 Qualcomm

In view of the importance of thiosemicarbazone derivatives, the present communication reports the synthesis of 4-formylpyridine N(4)-methyl thiosemicarbazone (H4FP4MT) with

Proposed work includes optimization using anomaly detection and removal algorithm and the security is enhanced by protecting the firewall files using the cryptographic algorithms..

(2014) A Role for the Mitochondrial Pyruvate Carrier as a Repressor of the Warburg Effect and Colon Cancer Cell Growth. (2008) Aldehyde Dehydrogenase as a Marker for